From Daughter to Researcher: Transforming Parkinson’s Through AI
Every person with Parkinson’s has a different story, AI may help us understand each one a little better
Like many researchers, my journey into Parkinson’s disease began with a personal connection. Growing up with a father with Parkinson’s, I learned early that a diagnosis is only the beginning of the story.
Parkinson’s disease is a progressive neurodegenerative condition best known for its effects on movement, but it can also affect mental health through changes in mood, thinking, sleep and perception. Perhaps most importantly, no two people experience it in exactly the same way. Symptoms, disease progression, and response to treatment can vary remarkably from one individual to another. My father’s Parkinson’s was one version of the disease, but there was no personalised blueprint. Neither is there for anyone else.
Today, I am a PhD candidate at King’s College London, where I use artificial intelligence (AI) to investigate Parkinson’s risk, psychiatric symptoms, disease progression, and treatment response. At its core, my research asks a simple question: how can we move beyond a one-size-fits-all approach and towards understanding Parkinson’s at the level of the individual?
When A Personal Question Became Scientific
For my father, Parkinson’s did not unfold in a straight line. It felt more like climbing terrain that kept changing: one section marked by movement symptoms, another by changes in mood, thinking or perception.
The terrain differs for every person, but every route is shaped by the same difficult truth: there is currently no cure, so treatment focuses on managing symptoms, and preserving quality of life.
For our family, navigating that terrain became a shared mission. Treatment often felt like an exercise in trial and error, with every decision involving trade-offs. For example, levodopa, the main medication used to improve movement symptoms, can also contribute to psychiatric side effects, including hallucinations and delusions. It could sometimes feel like solving one problem only to create another. Just when we thought we had found our footing, the terrain shifted again.
Long before I had the scientific language for it, I was already living with the question that would later shape my research: could those decisions be made with more foresight and less trial and error?
Turning Experience Into A Research Programme
My father lost his battle with Parkinson’s in June 2023, one month before I moved to London to begin my Master’s at King’s College London. Exactly one year later, in June 2024, I was accepted into the DRIVE-Health PhD programme with a proposal I had developed on using AI to transform Parkinson’s research and care. What began as a need to understand one person’s experience became a research programme focused on making Parkinson’s care more precise, predictive, and personalised.
I am now a PhD candidate in the Artificial Intelligence in Mental Health Lab at King’s, where I use machine learning, a form of AI that learns patterns from data, to investigate psychiatric outcomes, disease progression and treatment response in Parkinson’s. Alongside my PhD at King’s, I am also a Research Fellow at the Martinos Center for Biomedical Imaging in Boston where I use AI and neuroimaging to study Parkinson’s disease and treatment outcomes. Across my research, the aim is the same: to move beyond what happens on average and towards understanding risk, progression and treatment response at the level of the individual.
How AI Can Move Us Beyond Trial and Error
Parkinson’s is not a one-size-fits-all diagnosis. Clinicians already adapt care to each person, but much of the evidence available to them still tells us what works on average.
AI can help by looking at many pieces of information at once. A machine-learning model can combine clinical assessments, cognitive tests, genetic information and brain imaging, then identify patterns that may be difficult to see when each measure is considered alone. The goal is not for an algorithm to dictate care, but to provide evidence that supports better-informed, more personalised decisions.
Can we identify who is at greater risk of psychiatric symptoms? Can we recognise patterns linked to faster progression? Can we better estimate who may benefit from a treatment and who may experience side effects?
Prediction is only useful if it improves what happens next. Used well, AI could support earlier and more accurate diagnosis, identify emerging complications sooner, estimate which treatments are most likely to help a particular person, and anticipate who may be more vulnerable to side effects. That could mean earlier, better-targeted treatment with fewer cycles of adjustment. The promise of personalisation is not perfect certainty, but a more informed starting point than trial and error.
How I Use AI to Personalise Parkinson’s
At King’s, my research focuses on prediction at two different stages of Parkinson’s. The first one is recognition and diagnosis. The changes associated with Parkinson’s can begin more than a decade before diagnosis, and even after movement symptoms emerge, around a year can pass before someone receives a diagnosis in the UK.
Using routinely available clinical information, I am developing an AI model to estimate an individual’s risk of developing Parkinson’s earlier. The aim is to make early prediction as feasible as possible, opening the door to closer monitoring, earlier access to clinical trials and, as disease-modifying treatments emerge, interventions designed to delay onset or progression.
I also focus on Parkinson’s psychosis. Mental health difficulties are common in Parkinson’s, and more than half of people may experience psychosis during the course of the condition. It is one of its most serious complications, associated with care-home placement and poorer survival, yet there is currently no routine way to predict who will develop it. In my work on predicting Parkinson’s psychosis, I combine machine learning with brain imaging to identify patterns linked to psychosis and, ultimately, improve individual risk prediction.
At the Martinos Center for Biomedical Imaging in Boston, I apply the same principle to treatment decisions. Deep brain stimulation is a surgical treatment that can improve movement symptoms when medication no longer controls them well, but its effects are not identical for everyone. Using clinical and brain-imaging data, I investigate whether AI can help predict who may benefit and what outcomes might realistically be expected after surgery. The goal is to make a major treatment decision more individual: not simply asking whether deep brain stimulation works, but what it is likely to mean for the person considering it.
What Lived Experience Brings to Science
Watching my father’s Parkinson’s change over the years showed me how many forms one diagnosis can take. The symptoms that dominated at one stage could give way to different challenges at another, including changes in movement, cognition or mental health.
People with Parkinson’s may share a diagnosis, but the course of the condition, the difficulties they face, and the way they respond to treatment can look very different.
That perspective keeps me focused on building tools that preserve the individual within the data. Personalisation means asking not only what happens on average, but what may happen for this person, at this stage, and which information could support a better-informed decision.
I know there are people with Parkinson’s, caregivers, and family members navigating their own changing terrain now, searching for answers about symptoms, treatment and what comes next. I can only hope that my research gives them answers I once searched for, or that this story gives someone the confidence to become the researcher who finds the next one. Sometimes a scientific career begins long before a laboratory, with a question that matters too much to leave unanswered.
